Accuracy assessment of pharmacogenetically predictive warfarin dosing algorithms in patients of an academic medical center anticoagulation clinic

UMMS Affiliation

Department of Medicine, Division of Cardiovascular Medicine; Department of Medicine, Division of Preventive and Behavioral Medicine

Publication Date


Document Type



*Academic Medical Centers; Aged; Algorithms; Anticoagulants; Aryl Hydrocarbon Hydroxylases; Blood Coagulation; *Drug Dosage Calculations; Drug Monitoring; Female; Genetic Testing; Genotype; Humans; International Normalized Ratio; Male; Massachusetts; Middle Aged; Mixed Function Oxygenases; *Outpatient Clinics, Hospital; *Pharmacogenetics; Phenotype; Reproducibility of Results; Retrospective Studies; Warfarin


Behavioral Disciplines and Activities | Behavior and Behavior Mechanisms | Community Health and Preventive Medicine | Preventive Medicine


The objectives of this retrospective cohort study are to evaluate the accuracy of pharmacogenetic warfarin dosing algorithms in predicting therapeutic dose and to determine if this degree of accuracy warrants the routine use of genotyping to prospectively dose patients newly started on warfarin. Seventy-one patients of an outpatient anticoagulation clinic at an academic medical center who were age 18 years or older on a stable, therapeutic warfarin dose with international normalized ratio (INR) goal between 2.0 and 3.0, and cytochrome P450 isoenzyme 2C9 (CYP2C9) and vitamin K epoxide reductase complex subunit 1 (VKORC1) genotypes available between January 1, 2007 and September 30, 2008 were included. Six pharmacogenetic warfarin dosing algorithms were identified from the medical literature. Additionally, a 5 mg fixed dose approach was evaluated. Three algorithms, Zhu et al. (Clin Chem 53:1199-1205, 2007), Gage et al. (J Clin Ther 84:326-331, 2008), and International Warfarin Pharmacogenetic Consortium (IWPC) (N Engl J Med 360:753-764, 2009) were similar in the primary accuracy endpoints with mean absolute error (MAE) ranging from 1.7 to 1.8 mg/day and coefficient of determination R (2) from 0.61 to 0.66. However, the Zhu et al. algorithm severely over-predicted dose (defined as >or=2x or >or=2 mg/day more than actual dose) in twice as many (14 vs. 7%) patients as Gage et al. 2008 and IWPC 2009. In conclusion, the algorithms published by Gage et al. 2008 and the IWPC 2009 were the two most accurate pharmacogenetically based equations available in the medical literature in predicting therapeutic warfarin dose in our study population. However, the degree of accuracy demonstrated does not support the routine use of genotyping to prospectively dose all patients newly started on warfarin.

DOI of Published Version



J Thromb Thrombolysis. 2010 Aug;30(2):220-5. Link to article on publisher's site

Journal/Book/Conference Title

Journal of thrombosis and thrombolysis

Related Resources

Link to Article in PubMed

PubMed ID